


Model Order reduction for seismic imaging
Funded by: | NWO - DeepNL - WP2 |
PhD: |
|
Supervisor: | |
Collaboration: | Prof.dr. Jeannot Tramper - UU, Prof.dr. Marie-Colette van Lieshout - UT, Dr. Hanneke Paulssen - UU |
total project:
Workpackage P2: Model Order reduction for seismic imaging:
Monitoring seismicity via full-waveform inversion is computationally expensive and challenging, as the high-dimensional discretization consists of several millions of degrees of freedom and we need to find the velocity and elastic structure in every element of the high-dimensional discretization; the latter makes the parameter space very large. To speed up the calculations, Model Order Reduction (MOR) can be performed. However, due to the large-scale nature of the problem, classical MOR methods become unfeasible, as they would require prohibitively large reduced spaces. Moreover, classical MOR methods are not well-suited for wave propagation problems due to slow convergence and instabilities. The goal of this project is to develop MOR methods that are well-suited for speeding up full-waveform inversion, overcoming the above-mentioned challenges.